SmartBin

ThinkCity Hackathon Project Team: Yasmin Sikavi, Kai McCormick, Xavier Gilmour, Jooyoung Park, Yujeong An

Inspiration

The core problem is that people are frequently confused about where to dispose of their waste—whether it belongs in trash, recycling, or compost. This confusion leads to items being placed in the wrong containers, causing significant cross-contamination. This issue is critical because a 2024 study by the Recycling Partnership found that only 21% of residential recycling is actually recycled. We wanted to create a solution that removes the guesswork for the user and improves urban hygiene.

What it does

SmartBin is an all-in-one disposal solution where users can place all types of waste into a single large bin. The system uses Gemini AI to automatically detect and classify each item. Once the disposal method is identified, a mechanical system—functioning like a conveyor belt—drops the item into the correct internal compartment. To streamline building management, when the bin is full, an integrated vacuum tube system transports the sorted waste directly to centralized building collection points.

How we built it

Hardware Brain: We used a Single Board Computer (SBC), specifically a Raspberry Pi, to manage on-device processing and hardware triggers.
Vision & AI: The SBC captures multiple photos of an object and sends them to Gemini’s vision model. The AI classifies the item into one of four categories: Recycle, Trash, Compost, or Reject.
Cloud Infrastructure: We utilized a cloud-based setup including an S3 Bucket to cache object data and store images for analysis.
Automation: The system monitors the fullness status of internal compartments and controls the mechanical sorting and vacuum transport systems.

Impact & KPIs

Efficiency: SmartBin helps reduce cross-contamination, which lowers processing costs and prevents waste disposal facilities from rejecting entire loads.
Hygiene: By eliminating overflowing bins, the system reduces pests and improves overall city health.
Target Performance: Our goal is to achieve an 85% classification accuracy and a minimum 50% total waste diversion rate.

What's next for SmartBin

Consumer Scaling: Developing a smaller version of the device for at-home use.
Advanced Processing: Adding a cleaning feature for dirty jars and cans to increase their eligibility for recycling.
Optimization: Implementing trash compression technology to maximize the bin's capacity before it needs to be emptied.
Expansion: Deployment in high-traffic areas such as offices, college campuses, and retail stores.